Symmetry, Integrability and Geometry: Methods and Applications
SIGMA 4 (2008), 031, 18 pages
Recent Applications of the Theory of Lie Systems
in Ermakov Systems?
José F. CARIÑENA, Javier DE LUCAS and Manuel F. RAÑADA
Department of Theoretical Physics, University of Zaragoza, 50.009 Zaragoza, Spain
E-mail: jfc@unizar.es, dlucas@unizar.es, mfran@unizar.es
Received November 02, 2007, in final form February 04, 2008; Published online March 12, 2008
Original article is available at http://www.emis.de/journals/SIGMA/2008/031/
Abstract. We review some recent results of the theory of Lie systems in order to apply
such results to study Ermakov systems. The fundamental properties of Ermakov systems,
i.e. their superposition rules, the Lewis–Ermakov invariants, etc., are found from this new
perspective. We also obtain new results, such as a new superposition rule for the Pinney
equation in terms of three solutions of a related Riccati equation.
Key words: superposition rule; Pinney equation; Ermakov systems
2000 Mathematics Subject Classification: 34A26; 34A05
1
Introduction
Nonlinear equations have been of an increasing interest in Physics during the last thirty years
and one of the simplest examples is the today called Milne–Pinney equation
ẍ = −ω 2 (t)x +
k
,
x3
(1)
where k a is real constant with values depending on the field in which the equation is to
be applied. Such equation was introduced by an Ukrainian mathematician of the nineteenth
century, V.P. Ermakov, as a way of looking for a first integral for the time-dependent harmonic
oscillator [1]. He was using some of Lie’s ideas for dealing with ordinary differential equations
with tools of classical geometry. Lie had obtained a characterisation of non-autonomous systems
of first-order differential equations
dxi
= Y i (t, x),
dt
i = 1, . . . , n,
(2)
admitting a superposition rule [2]. Such a problem has being receiving these last years very
much attention because of its very important applications in physics and mathematics, see [3,
4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17]. This approach has recently been revisited from a
more geometric perspective in [18] where the rôle of the superposition function is played by an
appropriate connection. This new approach allows us to consider a superposition of solutions
of a given system in order to obtain solutions of another system as it was done in [19]. Our
aim in this paper is to show how such a superposition rule may be understood from a geometric
viewpoint in some interesting cases, as the Milne–Pinney equation (1) [20, 21], the Ermakov
system and its generalisations.
Recall that Ermakov systems are systems of second-order differential equations composed
by the Milne–Pinney differential equation (1) together with the corresponding time-dependent
?
This paper is a contribution to the Proceedings of the Seventh International Conference “Symmetry in
Nonlinear Mathematical Physics” (June 24–30, 2007, Kyiv, Ukraine). The full collection is available at
http://www.emis.de/journals/SIGMA/symmetry2007.html
2
J.F. Cariñena, J. de Lucas and M.F. Rañada
harmonic oscillator. They have received lot of attention since its introduction in Ermakov’s
paper [1], where many of its important properties have been exhibited. For instance, he found the
time-dependent invariant called Ermakov–Lewis invariant. In a short paper Pinney [21] was able
to prove the fundamental result that the general solution of the Milne–Pinney equation can be
described in terms of an arbitrary pair of independent particular solutions of the corresponding
time-dependent harmonic oscillator and two constants.
Ermakov systems have also been broadly studied in Physics since its introduction until now.
They appear in the study of the Bose–Einstein condensates and cosmological models [22, 23, 24],
and in the solution of time-dependent harmonic or anharmonic oscillators [25, 26, 27, 28, 29].
Much work has also been devoted to use Hamiltonian or Lagrangian structures in the study of
such a system, see e.g. [30, 31], and many generalisations or new insights from the mathematical
point of view can be found in [32, 33, 34, 35, 36, 37, 38, 39, 40].
Our aim in this paper is to review the mathematical insight developed in [6] in order to
study the properties of Ermakov systems. With this objective, we review in Section 2 the
concept of superposition rule, a map describing the general solution of a certain kind of system
of differential equations called Lie systems, in terms of a finite set of particular solutions of it.
The concept is generalised in Section 3 to express the general solution of a system of second-order
differential equations (hereafter shortened as SODE) in terms of arbitrary particular solutions of
a new Lie system and some constants. We prove that the theory developed by Lie can easily be
adapted to deal with many examples of such SODE systems, and in this way the superposition
rules which are generally used for first-order differential equations can also be used for dealing
with Ermakov systems. We find room in this way for implicit nonlinear superposition rules
in the terminology of [36, 41]. The Ermakov–Lewis invariants [29] appear in a natural way as
functions defining the foliation associated to the superposition rule. Some simple examples are
used to illustrate the theory. Section 4 is devoted to describe the possibility of a superposition
of solutions of a system to give a solution of another related one, and we show how Pinney’s
results [21] fit in this framework, i.e. we recover the known superposition rule in which the
solutions for the Pinney equation are expressed in terms of two solutions of the corresponding
t-dependent harmonic oscillator and as a straightforward application of our method we show
that the general solution of the Pinney equation can also be written as a superposition of three
particular solutions of the Riccati equation.
2
Systems of dif ferential equations
admitting a superposition rule
A superposition rule for solutions of (2) is determined by a function Φ : Rn(m+1) → Rn ,
x = Φ(x(1) , . . . , x(m) ; k1 , . . . , kn ),
x(a) ∈ Rn ,
a = 1, . . . , m,
such that the general solution can be written, for sufficiently small t, as
x(t) = Φ(x(1) (t), . . . , x(m) (t); k1 , . . . , kn ),
where {x(a) (t) | a = 1, . . . , m} is a fundamental set of particular solutions of the system (2)
and k = (k1 , . . . , kn ) is a set of n arbitrary constants, whose values fix each particular solution.
A superposition rule provides us a method for finding the general solution of (2) in terms of
a set of particular solutions.
According to the Implicit Function Theorem, the function Φ : Rn(m+1) → Rn can be, at least
locally around generic points, inverted, so we can write
k = Ψ(x(0) , . . . , x(m) )
Recent Applications of the Theory of Lie Systems in Ermakov Systems
3
for a certain function Ψ : Rn(m+1) → Rn . Hereafter in order to handle a shorter notation
we started writing x(0) instead of x. The level sets of the superposition function Ψ define
a foliation that is invariant under permutations of the (m + 1) variables. The functions Φ and Ψ
are related by:
k = Ψ(Φ(x(1) , . . . , x(m) ; k1 , . . . , kn ), x(1) , . . . , x(m) ).
(3)
The fundamental property of the superposition function Ψ is that as
k = Ψ(x(0) (t), x(1) (t), . . . , x(m) (t)),
the function Ψ(x(0) , . . . , x(m) ) is constant along any (m + 1)-tuple of solutions of the system (2).
This implies that if the ‘diagonal prolongation’ Ye (t, x(0) , . . . , x(m) ) of the t-dependent vector
field Y (t, x) = Y i (t, x)∂/∂xi is defined by
Ye (t, x(0) , . . . , x(m) ) =
m
X
Ya (t, x(a) ),
t ∈ R,
a=0
where
Ya (t, x(a) ) =
n
X
Y i (t, x(a) )
i=1
∂
,
∂xi(a)
then Ye (t, x(0) , . . . , x(m) ) is a t-dependent vector field on Rn(m+1) tangent to the level sets of Ψ,
i.e. the components Ψi are constants of motion. The level sets of Ψ corresponding to regular
values define a n-codimensional foliation F on an open dense subset U ⊂ Rn(m+1) and the family
{Ye (t, ·), t ∈ R} of vector fields in Rn(m+1) consists of vector fields tangent to the leaves of this
foliation.
Remark that, as pointed out in [18], for any k = (k1 , . . . , kn ) ∈ Rn and each (x(1) , . . . , x(m) ) ∈
Rnm there is one point (x(0) , x(1) , . . . , x(m) ) on the level set Fk of the foliation F, namely, such
that (Φ(x(1) , . . . , x(m) ; k), x(1) , . . . , x(m) ) ∈ Fk (cf. (3)); then, the projection onto the last m
factors
pr : (x(0) , x(1) , . . . , x(m) ) ∈ Rn(m+1) 7→ (x(1) , . . . , x(m) ) ∈ Rnm
induces diffeomorphisms on the leaves Fk of F. Such a foliation gives us the superposition
principle without referring to the function Ψ: if we fix the point x(0) (0) and m solutions
x(1) (t), . . . , x(m) (t), then x(0) (t) is the unique point in Rn such that (x(0) (t), x(1) (t), . . . , x(m) (t))
belongs to the same leaf of F as (x(0) (0), x(1) (0), . . . , x(m) (0)). This means that only F really
matters for the superposition rule.
Lie’s main result [2] can be expressed as follows:
Lie’s Theorem. The system (2) on a differentiable manifold N admits a superposition rule if
and only if the t-dependent vector field Y (t, x) can be locally written in the form
Y (t, x) =
r
X
bα (t) Xα (x),
(4)
α=1
where the vector fields Xα , with α = 1, . . . , r, close on a r-dimensional real Lie algebra, i.e. there
exist r3 real numbers cαβ γ such that
[Xα , Xβ ] =
r
X
cαβ γ Xγ ,
∀ α, β = 1, . . . , r.
γ=1
In this case we say that (2) is a Lie system.
(5)
4
J.F. Cariñena, J. de Lucas and M.F. Rañada
The number m of solutions involved in the superposition rule for the Lie system defined
by (4) with generic bα (t) is the minimal k such that the diagonal prolongations of X1 , . . . , Xr
e1 , . . . , X
er , are linearly independent at (generically) each point: for k = m the only real
to N k , X
numbers solution of the linear system
r
X
eα (x(1) , . . . , x(k) ) = 0,
cα X
α=1
at a generic point (x(1) , . . . , x(k) ) is the trivial solution cα = 0, with α = 1, . . . , r, and there are
nontrivial solutions for k < m. Then, the superposition function Ψ is made up of t-independent
constants of motion for the prolonged t-dependent vector field Ye . We shall call them firstintegrals.
A possible generalisation consists on considering foliations of codimension different from n,
or even choosing different Lie systems with the same associated Lie algebra for defining the
prolonged vector field. These facts will be illustrated with several examples in the following
sections.
3
SODE Lie systems
A system of second-order differential equations
d2 xi
dx
i
= F t, x,
,
dt2
dt
i = 1, . . . , n
can be studied through the system of first-order differential equations
ẋi = v i ,
v̇ i = F i (t, x, v),
i = 1, . . . , n,
with an associated t-dependent vector field
X = vi
∂
∂
+ F i (t, x, v) i .
∂xi
∂v
We call SODE Lie systems those SODE for which X is a Lie system, i.e. X can be written as
a linear combination with t-dependent coefficients of vector fields closing a finite-dimensional
real Lie algebra. Even if we are focusing here our attention on second-order differential equations,
the approach can be straightforwardly applied for r-order differential equations. Some particular
examples of second-order Lie systems are given in next subsections.
3.1
Milne–Pinney equation
The Milne–Pinney equation (1) is a second-order nonlinear differential equation [20, 21] that
describes the time-evolution of an isotonic oscillator [42, 43] (also called pseudo-oscillator), i.e.
a harmonic oscillator with inverse quadratic potential [32]. This nonlinear oscillator shares
with the harmonic one the property of having a period independent of the energy [44], i.e. it is
a isochronous system, and its quantum version has an equispaced spectrum [45, 46].
Note that if x(t) is solution of (1) then x0 (t) = −x(t) is solution too and we can restrict
ourselves to consider solutions in the half-line R+ = {x ∈ R | x > 0}. Thus, we are interested in
the solutions of (1) that are curves in R+ .
Recent Applications of the Theory of Lie Systems in Ermakov Systems
5
We can relate the Milne–Pinney equation with a system of first-order differential equations
defined in TR+ by introducing a new auxiliary variable v ≡ ẋ. Such system is given by
ẋ = v,
v̇ = −ω 2 (t)x +
k
,
x3
(6)
where x ∈ R+ and (x, v) ∈ Tx R+ . By the sake of simplicity we assume from now on that k > 0
but similar results can be obtained for k < 0. Now, the associated t-dependent vector field
over TR+ is
∂
∂
k
2
X=v
+ −ω (t)x + 3
.
∂x
x
∂v
This is a Lie system because X can be written as
X = L2 − ω 2 (t)L1 ,
where the vector fields L1 and L2 are given by
L1 = x
∂
,
∂v
L2 = v
∂
k ∂
+
,
∂x x3 ∂v
which are such that
[L1 , L2 ] = 2L3 ,
[L3 , L2 ] = −L2 ,
[L3 , L1 ] = L1
with
1
L3 =
2
∂
∂
x
−v
∂x
∂v
,
i.e. they span a 3-dimensional real Lie algebra g isomorphic to sl(2, R). Another different
relation of the Lie algebra sl(2, R) with this system has been pointed out but in the context
of Lie symmetries [47, 48]. On the contrary here sl(2, R) is the Lie algebra of vector fields
of the Lie system and the search for the general solution of (6) can be reduced to finding
the solution starting from the neutral element of the equation in the group SL(2, R) given by
ġ g −1 = ω 2 (t)a1 − a2 (see [5, 6]), where {a1 , a2 , a3 } is the following basis of sl(2, R):
1 −1 0
0 0
0 −1
,
a3 =
,
(7)
,
a2 =
a1 =
−1 0
0 0
0 1
2
which satisfy the same commutation relations as the vector fields {Lj | j = 1, 2, 3}. Actually, it
is possible to show that Lj is the fundamental vector field corresponding to aj with respect to
the action Φ : (A, (x, v)) ∈ SL(2, R) × TR+ → (x̄, v̄) ∈ TR+ given by:
s
2
k + (βv + αx)(δv + γx) + k(δβ/x2 )
x̄ =
,
(δv + γx)2 + k(δ/x)2
s
kδ 2
x2
α β
2
v̄ = κ (δv + γx) + 2 1 − 2 2
if A ≡
,
γ δ
x
δ x̄
where κ is ±1 or 0 depending on the initial point (x, v) and the element of the group SL(2, R)
that acts on it. In order to obtain an explicit expression for κ in terms of A and (x, v) we can
use the following decomposition for any element of the group SL(2, R)
α /2
1 α2
1 0
e 3
0
A = exp(−α2 a2 ) exp(−α1 a1 ) exp(−α3 a3 ) =
,
0 1
α1 1
0
e−α3 /2
6
J.F. Cariñena, J. de Lucas and M.F. Rañada
from where we obtain that α1 = γδ and α2 = β/δ. As we know that Φ(exp(−α3 a3 ), (x, v)) is
the integral curve of the vector field L3 starting from the point (x, v) parametrised by α3 , it is
straightforward to check that
(x1 , v1 ) ≡ Φ(exp(−α3 a3 ), (x, v)) = (exp(α3 /2)x, exp(−α3 /2)v),
and in a similar way
(x2 , v2 ) ≡ Φ(exp(−α1 a1 ), (x1 , v1 )) = (x1 , α1 x1 + v1 ).
Finally, we want to obtain (x̄, v̄) = Φ(exp(−α2 a2 ), (x2 , v2 )), and taking into account that the
integral curves of L2 satisfy that
x3 dv
dx
=
= dα2 ,
k
v
(8)
it turns out that when k > 0 we have v̄ 2 + k/x̄2 = v22 + k/x22 ≡ λ with λ > 0. Thus, using this
fact and (8) we obtain
k 1/2 dv
= dα2 ,
(λ − v 2 )3/2
and integrating v between v2 and v̄,
λ
v2
1
v̄
= α2 1/2 +
= 1/2 (α2 λ + v2 |x2 |) .
2
1/2
1/2
2
(λ − v̄ )
k
(λ − v2 )
k
As κ = sign[v̄], we see that κ is given by
β
kδβ |x|
2
κ = sign[α2 λ + v2 |x2 |] = sign
(xγ + vδ) + 2 +
(vδ + xγ) .
δ
x
δ
Non-trivial first-integrals independent of ω(t) for the equation (1) do not exist, i.e. there is not
any t-independent constant of motion I : U ⊂ TR+ → R such that XI = 0 for any function ω(t).
This is equivalent to dI(Lj ) = 0 for Lj with j = 1, 2, 3. Thus, the integrals of motion we are
looking for verify that dI vanishes on the involutive distribution V(x) ' hL1 (x), L2 (x), L3 (x)i
generated by the fundamental vector fields Lj . In almost any point we have that V(x) = Tx TR+ ,
and as dI = 0 on dense subsets in TR+ , the only possibility is dI = 0, therefore I is a constant,
and the integral of motion is trivial.
3.2
Generalised Ermakov system
Consider a possible generalisations of the Ermakov system given by
1
f (y/x),
x3
1
ÿ = −ω 2 (t)y + 3 g(y/x).
y
ẍ = −ω 2 (t)x +
(9)
This system of differential equations has been broadly studied in [27, 30, 35, 37, 41, 49, 50]. In
this section we analyse how can we make use of the theory of Lie systems in order to obtain the
known first-integral of motion for this system of differential equations.
Note that if (x(t), y(t)) is a solution, then (−x(t), −y(t)) is also a solution for (9), and therefore
we can focus our attention on the solutions of generalised Ermakov system which are curves in
the manifold R+ × R.
Recent Applications of the Theory of Lie Systems in Ermakov Systems
7
The generalised Ermakov system (9) can be written as a system of first-order differential
equations with a double number of degrees of freedom by introducing two new variables vx
and vy ,
ẋ = vx ,
v̇x = −ω 2 (t)x +
1
f (y/x),
x3
ẏ = vy ,
v̇y = −ω 2 (t)y +
1
g(y/x),
y3
(10)
and its solutions can be studied from the integral curves of the t-dependent vector field X in
T(R+ × R) given by
∂
∂
1
∂
∂
1
X = vx
+ −ω 2 (t)y + 3 g(y/x)
,
+ vy
+ −ω 2 (t)x + 3 f (y/x)
∂x
∂y
x
∂vx
y
∂vy
which can be written as a linear combination
X = N2 − ω 2 (t)N1 ,
where N1 and N2 are the vector fields
N1 = x
∂
∂
+y
,
∂vx
∂vy
N2 = vx
∂
1
∂
∂
∂
1
+ vy
+ 3 f (y/x)
+ 3 g(y/x)
.
∂x
∂y x
∂vx y
∂vy
Note that these vector fields generate a 3-dimensional real Lie algebra with a third generator
∂
∂
∂
∂
1
x
+y
− vx
− vy
.
N3 =
2
∂x
∂y
∂vx
∂vy
In fact, as
[N1 , N2 ] = 2N3 ,
[N3 , N1 ] = N1 ,
[N2 , N3 ] = N2 ,
they generate a Lie algebra isomorphic to sl(2, R). Therefore the generalised Ermakov system
is a SODE Lie system with associated Lie algebra sl(2, R).
The integrable distribution associated to the Lie system (10) is of dimension three, while the
manifold T(R+ ×R) is 4-dimensional, and then there exists a first-integral of motion, F : R4 → R,
for any ω 2 (t). This first-integral F satisfies Ni F = 0 for i = 1, 2, 3, but as [N1 , N2 ] = 2N3 it
is enough to impose N1 F = N2 F = 0. Then, if N1 F = 0, by the method of characteristics we
can conclude that there exists a function F̄ : R3 → R such that F (x, y, vx , vy ) = F̄ (x, y, ξ) with
ξ = xvy − yvx ). The condition N2 F = 0 reads now
∂ F̄
∂ F̄
y
x
∂ F̄
vx
+ vy
+ − 3 f (y/x) + 3 g(y/x)
= 0.
∂x
∂y
x
y
∂ξ
We can therefore consider the associated system of the characteristics:
dx
dy
dξ
=
= y
vx
vy
− x3 f (y/x) +
and using that
−y dx + x dy
dx
dy
=
=
,
ξ
vx
vy
,
x
g(y/x)
y3
8
J.F. Cariñena, J. de Lucas and M.F. Rañada
we arrive to
−y dx + x dy
dξ
= y y
ξ
− x3 f ( x ) +
,
x
g( xy )
y3
i.e.
y2d
−
x
y
ξ
=
dξ
+
− xy3 f ( xy )
x
g( xy )
y3
and integrating we obtain the following first-integral
Z x/y 1 2
1
1
1
+ζg
dζ = C.
ξ +
− 3f
2
ζ
ζ
ζ
This first-integral allows us to determine, by means of quadratures, a solution of a subsystem
in terms of a solution of the other equation.
3.3
The harmonic oscillator with time-dependent frequency
Our aim in this section is to exhibit with a simple example that we can define new Lie systems
of differential equations admitting first-integrals of motion which do not depend on the timedependent coefficients by putting together some copies of the same Lie system in the way
indicated in Section 2. These integrals of motion allow us to obtain relations between different
solutions of the initial Lie systems and can be used to construct a superposition rule if enough
copies have been added.
The equation of motion of a 1-dimensional harmonic oscillator with time-dependent frequency
is ẍ = −ω 2 (t)x, which is a second-order differential equation whose solutions are curves in R.
We can alternatively consider the following system of first-order differential equations
ẋ = v,
v̇ = −ω 2 (t)x,
(11)
whose solutions are the integral curves in TR of the t-dependent vector field
X=v
∂
∂
− ω 2 (t)x .
∂x
∂v
Now X is a linear combination X = X2 − ω 2 (t)X1 , where X1 and X2 are vector fields in TR:
X1 = x
∂
,
∂v
X2 = v
∂
,
∂x
such that
[X1 , X2 ] = 2X3 ,
[X3 , X1 ] = X1 ,
[X2 , X3 ] = X2 ,
(12)
where X3 is the vector field given by
1
∂
∂
X3 =
x
−v
.
2
∂x
∂v
As a consequence of Lie’s Theorem we see that X defines a Lie system with associated Lie
algebra sl(2, R). Actually, the vector fields {Xα | α = 1, 2, 3} are fundamental vector fields
corresponding to the usual linear action of SL(2, R) on R2 ' TR given by
α β
x̄ = αx + βv,
v̄ = γx + δv
if
A≡
.
γ δ
Recent Applications of the Theory of Lie Systems in Ermakov Systems
9
Remark that the elements of the basis of sl(2, R) given by (7) close the same commutation
relations as the vector fields {Xα | α = 1, 2, 3} which turn out to be the associated fundamental
vector fields for the elements aα .
Non-trivial first-integrals independent of ω(t) for the system (11) do not exist, because the
fundamental vector fields generate the involutive distribution V ' hX1 , X2 , X3 i whose values
in each point of TR generate the corresponding tangent space in almost any point and then
as dI vanishes on V, therefore dI vanishes in a generic point. Thus I is a trivial first integral of
motion.
As it was said before, non-trivial integrals of motion are needed to develop the procedures
of integration indicated in [6], i.e. we should add copies of the same differential equation until
the linear subspace of the involutive distribution of the total system in a generic point has
a dimension smaller than the dimension of the manifold. In this way there is room for a nonconstant function I such that dI vanishes on the subspace of the distribution associated with
the extended system of differential equations.
Thus, we first consider a system with two copies of the same differential equation,
ẍ1 = −ω 2 (t)x1 ,
ẍ2 = −ω 2 (t)x2 ,
(13)
which corresponds to a 2-dimensional isotropic harmonic oscillator with a time-dependent frequency and is associated with the following system of first-order differential equations
ẋ1 = v1 ,
v̇1 = −ω 2 (t)x1 ,
ẋ2 = v2 ,
v̇2 = −ω 2 (t)x2 ,
(14)
whose solutions are the integral curves of the t-dependent vector field
X = v1
∂
∂
∂
∂
+ v2
− ω 2 (t)x1
− ω 2 (t)x2
,
∂x1
∂x2
∂v1
∂v2
which is a linear combination, X = X2 − ω 2 (t)X1 , with X1 and X2 being the vector fields
X1 = x1
∂
∂
+ x2
,
∂v1
∂v2
X2 = v1
∂
∂
+ v2
,
∂x1
∂x2
which are such that
[X1 , X2 ] = 2X3 ,
[X3 , X1 ] = X1 ,
[X2 , X3 ] = X2 ,
(15)
where the vector field X3 is defined by
1
∂
∂
∂
∂
x1
+ x2
− v1
− v2
.
X3 =
2
∂x1
∂x2
∂v1
∂v2
The system of differential equations (14) is therefore a Lie system, i.e. the t-dependent vector
field can be written as a linear combination with time-dependent coefficients of vector fields Xα
closing a finite dimensional Lie algebra. Such vector fields are diagonal prolongations of the
fundamental vector fields of the initial system of differential equation and satisfy the same
commutation relations (15) as (12). The subspace of the distribution associated for this Lie
system has a rank lower or equal to the dimension of the Lie algebra. The Lie algebra associated
with the t-dependent vector field X is also sl(2, R).
10
J.F. Cariñena, J. de Lucas and M.F. Rañada
Note that the system (11) did not admit non-trivial first-integral of motion independent
of ω(t), but the system (13) does admit a first-integral because the subspace of the distribution
generated by the fundamental vector fields has rank three in almost any point and the dimension
of the total manifold is four. Thus, there exists a function F such that dF is vanishes on the distribution generated by the Xα ’s. The function F (x1 , x2 , v1 , v2 ) is such that X1 F = 0 iff there exists
a function F̄ (x1 , x2 , ξ) with ξ = x1 v2 − x2 v1 , such that F (x1 , x2 , v1 , v2 ) = F̄ (x1 , x2 , ξ), and then
the second condition X2 F = 0 implies that F̄ can only depend on ξ, i.e. F (x1 , x2 , v1 , v2 ) = F̂ (ξ).
From the commutation relation 2X3 = [X1 , X2 ], we see that the conditions X1 F = X2 F = 0
imply X3 F = 0, and therefore any first integral is a function of F (x1 , x2 , v1 , v2 ) = x1 v2 − x2 v1 ,
which physically corresponds to the angular momentum. This first integral can be seen from
the mathematical viewpoint as a partial superposition rule [6]. Actually, if x1 (t) is a solution
of the first equation, then we obtain for each real number k the first-order differential equation
for the variable x2
x1 (t)
dx2
= k + ẋ1 (t)x2 ,
dt
from where x2 can be found to be given by
Z t
dζ
0
,
x2 (t) = k x1 (t) + kx1 (t)
2
x1 (ζ)
where k 0 is a new integration constant.
In order to look for a superposition rule we should consider a system of some copies of (11)
with at least as many integrals of motion as the dimension of the initial manifold. Then it may be
possible to obtain the variables of the initial manifold explicitly in terms of the other variables.
Following the development of the Section 2 a way to obtain the number of necessary particular
solutions to obtain a superposition rule is to consider a set of copies of the initial system and
check out if the prolongations of the vector fields X1 , X2 and X3 are linearly independent in
a generic point.
In the case of two copies of the t-dependent harmonic oscillator it is possible to show that in
a generic point if λ1 X̃1 + λ2 X̃2 + λ3 X̃3 vanishes then λ1 = λ2 = λ3 = 0, therefore m = 2 and
consequently there is a superposition rule involving two particular solutions. In order to obtain
the superposition rule we need to consider three copies of the t-dependent harmonic oscillator
and study the system of first-order differential equations
ẋ1 = v1 ,
v̇1 = −ω 2 (t)x1 , ẋ2 = v2 ,
v̇2 = −ω 2 (t)x2 ,
ẋ = v,
v̇ = −ω 2 (t)x,
whose solutions are the integral curves of the t-dependent vector field
X = v1
∂
∂
∂
∂
∂
∂
+ v2
+v
− ω 2 (t)x1
− ω 2 (t)x2
− ω 2 (t)x ,
∂x1
∂x2
∂x
∂v1
∂v2
∂v
which is a linear combination, X = X2 − ω 2 (t)X1 , with X1 and X2 being the vector fields
X1 = x1
∂
∂
∂
+ x2
+x ,
∂v1
∂v2
∂v
X2 = v1
∂
∂
∂
+ v2
+v ,
∂x1
∂x2
∂x
which are such that
[X1 , X2 ] = 2X3 ,
[X3 , X1 ] = X1 ,
[X2 , X3 ] = X2 ,
Recent Applications of the Theory of Lie Systems in Ermakov Systems
11
where the vector field X3 is defined by
1
∂
∂
∂
∂
∂
∂
X3 =
x1
+ x2
+x
− v2
−v
.
− v1
2
∂x1
∂x2
∂x
∂v1
∂v2
∂v
The first-integrals F are as in the last case the solutions of X1 F = X2 F = 0, because
2X3 = [X1 , X2 ]. The condition X1 F = 0 says that there exists a function F̄ : R5 → R2
such that F (x1 , x2 , x, v1 , v2 , v) = F̄ (x1 , x2 , x, ξ1 , ξ2 ) with ξ1 (x1 , x2 , x, v1 , v2 , v) = xv1 − x1 v and
ξ2 (x1 , x2 , x, v1 , v2 , v) = xv2 −x2 v, and when written in terms of F̄ the condition X2 F = 0 implies
that F̄ is an arbitrary function of ξ1 and ξ2 (of course, ξ = x1 v2 − x2 v1 is also a first-integral).
They produce a superposition rule, because from
xv2 − x2 v = k1 ,
x1 v − v1 x = k2
we obtain the expected superposition rule for two solutions:
x = c1 x1 + c2 x2 ,
4
v = c1 v1 + c2 v2 ,
ci =
k1
,
k
k = x1 v2 − x2 v1 .
Construction of mixed superposition rules
In the preceding sections we have considered SODE Lie systems together with some copies of
the same system. In this section we show a way to deal with sets of maybe different systems of
differential equations sharing the same associated Lie algebra as Lie systems. These sets satisfy
certain conditions to be explained later on and can be used to obtain mixed superposition rules.
Suppose that a Lie system is given
r
dx X
=
bα (t)Xα (x),
dt
α=1
where the Xα with α = 1, . . . , r, close a finite-dimensional Lie algebra as given by (5). We want to
obtain a mixed superposition rule for this Lie system, i.e. a function Φ : N 1 ×· · ·×N m ×Rn → N
in such a way that any solution integral curve x(t) of X(t, ·) be given by
x(t) = Φ(x(1) (t), . . . , xm (t), k1 , . . . , kn ),
where x(1) (t), . . . , x(m) (t), are integral curves of the t-dependent vector fields on the corresponding manifolds N j with j = 1, . . . , m and given by
X
(1)
(t, ·) =
r
X
bα (t)Xα1 (·),
α=1
···························
r
X
bα (t)Xαm (·)
X (m) (t, ·) =
α=1
and in such a way that Xαj with j = 1, . . . , m, and α = 1, . . . , r, close the same commutation
relations as the vector fields Xα , i.e.
[Xαj , Xβj ] = cαβ γ Xγj ,
j = 1, . . . , m,
and
α, β = 1, . . . r.
The search for a set of vector fields closing a given Lie algebra in a certain manifold amounts
to look for different actions of the corresponding Lie group.
12
J.F. Cariñena, J. de Lucas and M.F. Rañada
Then, over Ñ we obtain that
[Xα , Xβj ] = 0,
[Xαj , Xβk ] = 0,
j, k = 1, . . . , m,
with j 6= k, and α, β = 1, . . . r,
and because of these commutation relations we can define in Ñ the vector fields
Ỹα = Xα +
m
X
Xαi
i=1
which satisfy the same commutation relations as the vector fields Xα in (5). Thus, the system
of differential equations determining the integral curves of the t-dependent vector field
Ỹ (t, ·) =
r
X
bα (t)Ỹα (·)
α=1
is a Lie system with the same associated Lie algebra as the initial system of differential equations.
Nevertheless, the dimension of the manifold is this time larger than the dimension of N because
we have put together several manifolds. The distribution Ṽ associated to this Lie system is
given by Ṽ(x̃) = hỸ1 (x̃), . . . , Ỹr (x̃)i and is involutive. The dimension of the subspace in a point
is lower or equal to r and therefore when the dimension of the manifold Ñ is larger than r there
are integrals of motion for any value of the time-dependent coefficients which may be used to
obtain superposition rules.
In the next section we give some examples of the application of this procedure to the Ermakov
system in order to both recover previously known properties and to find also new results.
4.1
Ermakov system
Consider the system of ordinary first-order differential equations [28, 40]
ẋ = vx ,
v̇x = −ω 2 (t)x,
ẏ = vy ,
v̇y = −ω 2 (t)y +
1
y3
made up by a 1-dimensional harmonic oscillator and the Milne–Pinney equation with k = 1,
which has the above-mentioned conditions. Its solutions are the integral curves of the t-dependent vector field
∂
∂
∂
1
∂
2
2
X = vx
+ vy
− ω (t)x
+ −ω (t)y + 3
,
∂x
∂y
∂vx
y
∂vy
which is a linear combination with time-dependent coefficients, X = X2 − ω 2 (t)X1 , of
X1 = x
∂
∂
+y
,
∂vx
∂vy
X2 = vx
∂
∂
1 ∂
+ vy
+
.
∂x
∂y y 3 ∂vy
This system closes on a sl(2, R) Lie algebra with X3 given by
∂
∂
1
∂
∂
X3 =
+y
− vx
− vy
x
.
2
∂x
∂y
∂vx
∂vy
The generators of this Lie system span a distribution of dimension two and there is no
first-integral of the motion for such subsystem. By adding the other sl(2, R) linear Lie system
Recent Applications of the Theory of Lie Systems in Ermakov Systems
13
appearing in the Ermakov system, the harmonic oscillator with the same time-dependent angular
frequency, as the rank of the space of the distribution in the 4-dimensional space is three, there is
an integral of motion. The first-integral can be obtained from X1 F = X2 F = 0. But X1 F = 0
means that there exists a function F̄ : R3 → R such that F (x, y, vx , vy ) = F̄ (x, y, ξ), with
ξ = xvy − yvx , and then X2 F = 0 is written
vx
∂ F̄
∂ F̄
x ∂ F̄
+ vy
+ 3
∂x
∂y
y ∂ξ
and we obtain the associated system of characteristics
xdy − ydx
y 3 dξ
d(x/y) ydξ
=
=⇒
+
= 0,
ξ
x
ξ
x
from where the following first-integral is found [29]:
2
2
x
x
2
ψ(x, y, vx , vy ) =
+ξ =
+ (xvy − yvx )2 ,
y
y
which is the well-known Ermakov–Lewis invariant [27, 28, 40].
4.2
The Pinney equation revisited
We mentioned before the possibility of obtaining solutions of a given system from particular
solutions of another related system. We next study a particular example which allows us to
recover the results obtained by Pinney long time ago [21]. Consider the system of first-order
differential equations:
ẋ = vx ,
ẏ = vy ,
ż = vz ,
v̇x = −ω 2 (t)x +
k
,
x3
v̇y = −ω 2 (t)y,
v̇z = −ω 2 (t)z,
which corresponds to the vector field
∂
∂
∂
k ∂
∂
∂
∂
2
X = vx
+ vy
+ vz
+
− ω (t) x
+y
+z
.
∂x
∂y
∂z x3 ∂vx
∂vx
∂vy
∂vz
The t-dependent vector field X can be expressed as X = N2 − ω 2 (t)N1 where N1 and N2 are
N1 = y
∂
∂
∂
+x
+z
,
∂vy
∂vx
∂vz
N2 = vx
∂
∂
∂
1 ∂
+ vy
+ vz
+ 3
,
∂x
∂y
∂z x ∂vx
These vector fields generate a 3-dimensional real Lie algebra with the vector field N3 given by
1
∂
∂
∂
∂
∂
∂
N3 =
x
+y
+z
− vx
− vy
− vz
.
2
∂x
∂y
∂z
∂vx
∂vy
∂vz
In fact, they generate a Lie algebra isomorphic to sl(2, R) because
[N1 , N2 ] = 2N3 ,
[N3 , N1 ] = N1 ,
[N2 , N3 ] = N2 .
14
J.F. Cariñena, J. de Lucas and M.F. Rañada
The dimension of the distribution generated by these fundamental vector fields is three
and the manifold of the Lie system is of dimension six, then there are three time-independent
integrals of motion which turn out to be the Ermakov invariant I1 of the subsystem involving
variables x and y, the Ermakov invariant I2 of the subsystem involving variables x and z, and
the Wronskian W of the subsystem involving variables y and z. They define a foliation with
3-dimensional leaves. This foliation can be used for obtaining a superposition rule.
The Ermakov invariants read as
y 2 z 2 1
1
I1 =
(yvx − xvy )2 + k
,
I2 =
(xvz − zvx )2 + k
,
2
x
2
x
where I1 and I2 are non-negative constants and the Wronskian W is:
W = yvz − zvy .
We can obtain an explicit expression of x in terms of y, z and the three first integrals I1 , I2 , W :
√ 1/2
p
2
x=
.
I2 y 2 + I1 z 2 ± 4I1 I2 − kW 2 yz
|W |
Remark that here W is a constant fixed by the two independent particular solutions of the
time-dependent harmonic oscillator x1 (t) and x2 (t) and only I1 and I2 play the role of constants
in this superposition rule for the Milne–Pinney equation. This is not a surprising fact because
the Milne–Pinney equation is a second-order differential equation. Note also that the values
of I1 and I2 are non-negative constants but should be chosen such that x(0) be real.
This can be interpreted, as pointed out by Pinney [21], as saying that there is a superposition
rule allowing us to express the general solution of the Milne–Pinney equation in terms of two
independent solutions of the corresponding harmonic oscillator with the same time-dependent
angular frequency.
4.3
A new superposition rule for the Pinney equation
A new mixed superposition rule involving solutions of a Riccati for obtaining the general solution of Pinney equations is obtained in this section as a straightforward application of our
development.
The t-dependent Riccati equation
dx
= b0 (t) + b1 (t)x + b2 (t)x2 .
dt
(16)
has been studied in [51, 52] from the perspective of the theory of Lie systems. We follow
here a very similar approach. From the geometric viewpoint the Riccati equation (16) can
be considered as a differential equation determining the integral curves of the time-dependent
vector field
Γ = (b0 (t) + b1 (t)x + b2 (t)x2 )
∂
,
∂x
which is a linear combination with time-dependent coefficients of the three vector fields
L1 =
∂
,
∂x
L2 = −x2
∂
,
∂x
L3 = −x
∂
,
∂x
which close on a 3-dimensional real Lie algebra, with defining relations
[L1 , L2 ] = 2L3 ,
[L3 , L1 ] = L1 ,
[L2 , L3 ] = L2 ,
(17)
therefore isomorphic to sl(2, R), because the commutation relations (17) are the same as (12).
Recent Applications of the Theory of Lie Systems in Ermakov Systems
15
The following particular case of Riccati equation:
dx
= −ω 2 (t) − x2 ,
dt
is the equation for the integral curves of the t-dependent vector field L = L2 − ω 2 (t)L1 . Consequently, we can apply the procedure of the Section 4 and consider the following system of
differential equations
ẋ1 = −ω 2 (t) − x21 ,
ẋ2 = −ω 2 (t) − x22 ,
ẋ3 = −ω 2 (t) − x23 ,
ẋ = v,
v̇ = −ω 2 (t)x +
k
x3
described by a vector field in R3 × TR+ , where (x1 , x2 , x3 ) ∈ R3 , x ∈ R+ and (x, v) ∈ Tx R+ .
According to our general recipe, consider the following vector fields
∂
∂
∂
∂
∂
∂
∂
∂
k ∂
+
+
+x ,
M2 = −x21
− x22
− x23
+v
+
,
∂x1 ∂x2 ∂x3
∂v
∂x1
∂x2
∂x3
∂x x3 ∂v
∂
∂
∂
1
∂
∂
M3 = −x1
− x2
− x3
+
x
−v
,
∂x1
∂x2
∂x3 2
∂x
∂v
M1 =
that, by construction, verify the same commutation relations as before, i.e.
[M1 , M2 ] = 2M3 ,
[M3 , M1 ] = M1 ,
[M3 , M2 ] = −M2 ,
and the full system of differential equations can be understood as the system of differential
equations for the determination of the integral curves of the t-dependent vector field M =
M2 − ω 2 (t)M1 . The dimension of the distribution associated to this Lie system is three in
almost any point and then there exist two integrals of motion. As 2M3 = [M1 , M2 ], it is enough
to find the simultaneously integrals of motion of M1 and M2 , i.e. a function F : R5 → R such
that M1 F = M2 F = 0.
Let us first look for first-integrals independent of x3 , i.e. we suppose that F depends just
on x1 , x2 , x and v. Using the method of characteristics, the condition M1 F = 0 means that
there is a function F̄ : R3 → R such that F (x1 , x2 , x, v) = F̄ (I1 , I2 , I3 ) with I1 , I2 and I3 given by
I1 = x1 − x2 ,
I2 = x2 − v/x,
I3 = x.
Then, the condition M2 F̄ = 0 reads in terms of the variables I1 , I2 , I3 and I4 ≡ v as
∂ F̄
I1 ∂ F̄
I2 ∂ F̄
∂ F̄
k ∂ F̄
2
2
I4 −2
−2
+
+ −I1 − 2I1 I2
− I2 + 4
= 0.
I3 ∂I1
I3 ∂I3 ∂I3
∂I1
I3 ∂I2
Thus, the linear term in I4 and the other one must vanish independently. The method of
characteristics applied to the first term means that there exists a function Fb : R2 → R such
that F̄ (I1 , I2 , I3 ) = Fb(K1 , K2 ) where
K1 =
I1
,
I2
K2 = I2 I32 .
Finally, taking into account the last result in M2 F̂ = 0 we obtain
b
kK1 ∂ Fb
k
∂F
2
−K1 − K1 + 2
− K2 +
= 0,
K2 ∂K2
K2 ∂K1
16
J.F. Cariñena, J. de Lucas and M.F. Rañada
and by means of the method of characteristics
1
K12 + K1 − kK
dK1
K22
=
.
dK2
K2 + Kk2
Finally, taking into account the last result we obtain the first-integral
C1 = K2 +
k + K22
,
K1 K2
that in terms of the initial variables reads as
v 2 k + (x2 − xv )2 x4
C1 = x2 −
.
x +
x
(x1 − x2 )x2
If we repeat the last procedure but under the assumption that the integral does not depend
on x2 we obtain the following first-integral:
v 2 k + (x3 − xv )2 x4
C2 = x3 −
x +
.
x
(x1 − x3 )x2
It is a long but straightforward computation to check out that both are integrals of M1 , M2
and M3 . We can obtain now the general solution x of (1) x in terms of x1 , x2 , x3 , C1 , C2 as
s
(C1 (x1 − x2 ) − C2 (x1 − x3 ))2 + k(x2 − x3 )2
x=
,
(18)
(C2 − C1 )(x2 − x3 )(x2 − x1 )(x1 − x3 )
where C1 and C2 are constants such that once x1 (t), x2 (t) and x3 (t) have been fixed they
make x(0) given by (18) to be real.
Thus we have obtained a new mixed superposition rule which allows us to express the general
solution of the Pinney equation in terms of three solutions of Riccati equations, and of course
two constants which determine each particular solution.
5
Conclusions and outlook
In this paper we have reviewed the theory developed in [6] in order to deal with a certain kind of
systems second-order differential equations that can be studied as Lie systems, in particular with
the Milne–Pinney equation and the Ermakov system. We have not only recovered in this way
some known results about these differential equations, i.e. integrals of motion or superposition
rules, but we have also been able to show new ones as a superposition rule for three solutions of
a Riccati equation to give us the general solution of the Milne–Pinney equation.
New applications of this formalism to relate a Milne–Pinney equation with a frequency ω(t)
with a t-dependent harmonic oscillator with different frequency ω 0 (t) related with ω(t) by means
of certain relations will be developed in forthcoming papers. In this way a new time-dependent
superposition rule is obtained and by means of it we can express the solution of a Milne–Pinney
equation with frequency ω(t) in terms of solutions of a time-dependent harmonic oscillator with
frequency ω 0 (t).
Another remarkable point to be studied is the extension of this formalism for studying some
generalised Ermakov systems as those appearing for instance in [38] and [39].
Acknowledgements
Partial financial support by research projects MTM2006-10531 and E24/1 (DGA) are acknowledged. JdL also acknowledge a F.P.U. grant from Ministerio de Educación y Ciencia and
a special grant from the Network of Mechanics, Geometry and Control.
Recent Applications of the Theory of Lie Systems in Ermakov Systems
17
References
[1] Ermakov V.P., Second-order differential equations. Conditions of complete integrability, Univ. Isz. Kiev
Series III 9 (1880), 1–25 (translation by A.O. Harin).
[2] Lie S., Vorlesungen über continuierliche Gruppen mit Geometrischen und anderen Anwendungen, edited
and revised by G. Scheffers, Teubner, Leipzig, 1893.
[3] Ibragimov N.H., An ABC of group analysis, Novoe v Zhizni, Nauke, Tekhnike. Seriya Matematika, Kibernetika, no. 8, Znanie, Moscow, 1989 (in Russian).
Ibragimov N.H., Introduction to modern group analysis, Tau, Ufa, 2000 (revised edition in English).
[4] Ibragimov N.H., Elementary Lie group analysis and ordinary differential equations, J. Wiley, Chichester,
1999.
[5] Winternitz P., Lie groups and solutions of nonlinear differential equations, in Nonlinear Phenomena, Editor
K.B. Wolf, Lecture Notes in Physics, Vol. 189, Springer-Verlag, New York, 1983, 263–305.
[6] Cariñena J.F., Grabowski J., Marmo G., Lie–Scheffers systems: a geometric approach, Bibliopolis, Napoli,
2000.
[7] Cariñena J.F., Grabowski J., Ramos A., Reduction of time-dependent systems admitting a superposition
principle, Acta Appl. Math. 66 (2001), 67–87.
[8] Cariñena J.F., Grabowski J., Marmo G., Some applications in physics of differential equation systems
admitting a superposition rule, Rep. Math. Phys. 48 (2001), 47–58.
[9] Anderson R.L., A nonlinear superposition principle admitted by coupled Riccati equations of the projective
type, Lett. Math. Phys. 4 (1980), 1–7.
[10] Harnad J., Winternitz P., Anderson R.L., Superposition principles for matrix Riccati equations, J. Math.
Phys. 24 (1983), 1062–1072.
[11] del Olmo M.A., Rodrı́guez M.A., Winternitz P., Simple subgroups of simple Lie groups and nonlinear
differential equations with superposition principles, J. Math. Phys. 27 (1986), 14–23.
[12] Kevrekidis P.G., Drossinos Y., Nonlinearity from linearity: the Ermakov–Pinney equation revisited, Math.
Comput. Simulation 74 (2007), 196–202.
[13] Leach P.G.L., Karasu A., Nucci M.C., Andriopoulos K., Ermakov’s superintegrable toy and nonlocal symmetries, SIGMA 1 (2005), 018, 15 pages, nlin.SI/0511055.
[14] del Olmo M.A., Rodrı́guez M.A., Winternitz P., Superposition formulas for rectangular matrix Riccati
equations, J. Math. Phys. 28 (1987), 530–535.
[15] Cariñena J.F., Ramos A., Riccati equation, factorization method and shape invariance, Rev. Math. Phys.
12 (2000), 1279–1304, math-ph/9910020.
[16] Cariñena J.F., Ramos A., A new geometric approach to Lie systems and physical applications, Acta Appl.
Math. 70 (2002), 43–69, math-ph/0110023.
[17] Cariñena J.F., Marmo G., Nasarre J., The nonlinear superposition principle and the Wei–Norman method,
Internat. J. Modern Phys. A 13 (1998), 3601–3627, physics/9802041.
[18] Cariñena J.F., Grabowski J., Marmo G., Superposition rules, Lie theorem and partial differential equations,
Rep. Math. Phys. 60 (2007), 237–258 math-ph/0610013.
[19] Cariñena J.F., de Lucas J., Rañada M.F., Nonlinear superposition rules and Ermakov systems, in Differential
Geometric Methods in Mechanics and Field Theory, Editors F. Cantrijn, M. Crampin and B. Langerock,
Academia Press, 2007, 15–33.
[20] Milne W.E., The numerical determination of characteristic numbers, Phys. Rev. 35 (1930), 863–67.
[21] Pinney E., The nonlinear differential equation y 00 + p(x)y + cy −3 = 0, Proc. Amer. Math. Soc. 1 (1950), 681.
[22] Hawkins R.M., Lidsey J.E., Ermakov–Pinney equation in scalar field cosmologies, Phys. Rev. D 66 (2002),
023523, 8 pages, astro-ph/0112139.
[23] Lidsey L.E., Cosmic dynamics of Bose–Einstein condensates, Classical Quantum Gravity 21 (2004), 777–785,
gr-qc/0307037.
[24] Haas F., Anisotropic Bose–Einstein condensates and completely integrable dynamical systems, Phys. Rev. A
65 (2002), 033603, 6 pages, cond-mat/0211353.
[25] Fernández Guasti M., Moya-Cessa H., Amplitude and phase representation of quantum invariants for the
time-dependent harmonic osicllator, Phys. Rev. A 67 (2003), 063803, 5 pages, quant-ph/0212073.
18
J.F. Cariñena, J. de Lucas and M.F. Rañada
[26] Gauthier S., An exact invariant for the time dependent double well anharmonic oscillators: Lie theory and
quasi-invariance groups, J. Phys. A: Math. Gen. 17 (1984), 2633–2639.
[27] Ray J.R., Reid J.L., More exact invariants for the time-dependent harmonic oscillator, Phys. Lett. A 71
(1979), 317–318.
[28] Dhara A.K., Lawande S.V., Time-dependent invariants and the Feynman propagator, Phys. Rev. A 30
(1984), 560–567.
[29] Lewis H.R., Classical and quantum systems with time-dependent harmonic-oscillator-type Hamiltonians,
Phys. Rev. Lett. 18 (1967), 510–512.
[30] Reid J.L., Ray J.R., Ermakov systems, Noether’s theorem and the Sarlet–Bahar method, Lett. Math. Phys.
4 (1980), 235–240.
[31] Cerveró J.M., Lejarreta J.D., Ermakov Hamiltonians, Phys. Lett. A 156 (1991), 201–205.
[32] Sarlet W., Exact invariants for time-dependent Hamiltonian systems with one degree of freedom, J. Phys. A:
Math. Gen. 11 (1978), 843–854.
[33] Govinder K.S., Athorne C., Leach P.G.L., The algebraic structure of generalized Ermakov systems in three
dimensions, J. Phys. A: Math. Gen. 26 (1993), 4035–4046.
[34] Athorne C., Rogers C., Ramgulam U., Osbaldestin A., On linearization of the Ermakov system, Phys.
Lett. A 143 (1990), 207–212.
[35] Sarlet W., Cantrijn F., A generalization of the nonlinear superposition idea for Ermakov systems, Phys.
Lett. A 88 (1982), 383–387.
[36] Reid J.L., Ray J.R., Ermakov systems, nonlinear superposition and solutions of nonlinear equations of
motion, J. Math. Phys. 21 (1980), 1583–1587.
[37] Govinder K.S., Leach P.G.L., Ermakov systems: a group theoretic approach, Phys. Lett. A 186 (1994),
391–395.
[38] Athorne C., Projective lifts and generalised Ermakov and Bernoulli systems, J. Math. Anal. Appl. 233
(1999), 552–563.
[39] Rogers C., Schief W.K., Bassom A., Ermakov systems with arbitrary order, dimension. Structure and
linearisation, J. Phys. A: Math. Gen. 29 (1996), 903–911.
[40] Leach P.G.L., Generalized Ermakov systems, Phys. Lett. A 158 (1991), 102–106.
[41] Sarlet W., Further generalization of Ray–Reid systems, Phys. Lett. A 82 (1981), 161–164.
[42] Calogero F., Solution of a three body problem in one dimension, J. Math. Phys. 10 (1969), 2191–2196.
[43] Perelomov A.M., Integrable systems of classical mechanics and Lie algebras, Birkhäuser Verlag, Basel, 1990.
[44] Chalykh O.A., Vesselov A.P., A remark on rational isochronous potentials, J. Nonlinear Math. Phys. 12
(2005), suppl. 1, 179–183, math-ph/0409062.
[45] Asorey M., Cariñena J.F., Marmo G., Perelomov A.M., Isoperiodic classical systems and their quantum
counterparts, Ann. Phys. 322 (2007), 1444–1465, arXiv:0707.4465.
[46] Cariñena J.F., Perelomov A.M., Rañada M.F., Isochronous classical systems and quantum systems with
equally spaced spectra, in Particles and Fields: Classical and Quantum, J. Phys. Conf. Ser. 87 (2007),
012007, 14 pages.
[47] Leach P.G.L., Karasu A., The Lie algebra sl(2, R) and so-called Kepler–Ermakov systems, J. Nonlinear
Math. Phys. 11 (2004), 269–275.
[48] Karasu A., Yildrim H., On the Lie symmetries of the Kepler–Ermakov systems, J. Nonlinear Math. Phys.
9 (2002), 475–482, math-ph/0306037.
[49] Ray J.R., Reid J.L., Exact time-dependent invariants for N -dimensional systems, Phys. Lett. A 74 (1979),
23–25.
[50] Ray J.R., Invariants for nonlinear equations of motion, Progr. Theoret. Phys. 65 (1981), 877–882.
[51] Cariñena J.F., Ramos A., Integrability of the Riccati equation from a group theoretical viewpoint, Internat.
J. Modern Phys. A 14 (1999), 1935–1951, math-ph/9810005.
[52] Cariñena J.F., de Lucas J., Ramos A., A geometric approach to integrability conditions for Riccati equations,
Electron. J. Differential Equations 2007 (2007), 122, 14 pages.